Yunhao (Andy) Ge


CS Ph.D. Candidate @ University of Southern California

Student Researcher @ Google Cloud AI

Department of Computer Science @ University of Southern California
Email: yunhaoge at usc dot edu

    [About Me] [News & Updates] [Publications] [Experience]

About Me


I am now a Ph.D. Candidate in the Department of CS at University of Southern California (USC) and a member of iLab, working with Prof. Laurent Itti. I'm also currently a student researcher at Google Cloud AI working with Dr. Sercan Arik and Dr. Jinsung Yoon . I also work closely with Dr. Ziyan Wu ( UII America ) and Dr. Vibhav Vineet ( Microsoft Research ).

Before that, I got my M.Sc. degree at Robotics Institute at Shanghai Jiao Tong University, and B.Eng. degree of Mechatronics at Shandong University. I have been lucky to work with Prof. Dinggang Shen and Dr. Shu Liao during my internship at United Imaging Intelligence, as well as Prof. Cewu Lu and Dr. Shuyun Chong when I was a Computer Vision Engineer at Flexiv Robotics.

I'm interested in Machine Learning, Computer vision, and their applications towards more Human-centric and Humanoid AI. My current research focuses include:

Research opportunities: I am happy to collaborate and/or answer questions about my research and USC academic programs. If you are interested, please send me an email. I especially encourage USC master students who want to involve exciting projects targeting top-tier conferences/journals to reach out.

News & Updates

Selected Publications [Google Scholar]

A Peek Into the Reasoning of Neural Networks: Interpreting with Structural Visual Concepts
TL;DR: Takes a step towards mimicking the reasoning process of NNs and provide logical, concept-level explanations for final model decisions.

Yunhao Ge, Yao Xiao, Zhi Xu, Meng Zheng, Srikrishna Karanam, Terrence Chen, Laurent Itti and Ziyan Wu
IEEE/ CVF International Conference on Computer Vision and Pattern Recognition (CVPR), 2021.

[paper] [github] [website] [video] [知乎] [机器之心] [AI科技评论]

Zero-shot Synthesis with Group-Supervised Learning
TL;DR: Using Controllable disentangled representation learning to simulate human knowledge factorization for imagination

Yunhao Ge, Sami Abu-El-Haija, Gan Xin and Laurent Itti
International Conference on Learning Representations (ICLR), 2021.

[paper] [code] [website] [Fonts Dataset] [USC Viterbi Press] [知乎] [AI科技评论]
[ USC News ] [ Tech Xplore ] [ Technology Networks ]

Encouraging Disentangled and Convex Representation with Controllable Interpolation Regularization
TL;DR: A simple yet efficient method: Controllable Interpolation Regularization (CIR), which creates a positive loop where the disentanglement and convexity can help each other

Yunhao Ge, Zhi Xu, Yao Xiao, Gan Xin, Yunkui Pang and Laurent Itti
arXiv, 2021.


Graph Autoencoder for Graph Compression and Representation Learning
TL;DR: Multi-kernel Inductive Attention Graph Autoencoder (MIAGAE) utilizes the node similarity and graph structure to compress all nodes and edges as a whole.

Yunhao Ge*, Yunkui Pang*, Linwei Li and Laurent Itti (*=equal contribution)
Neural Compression: From Information Theory to Applications--Workshop@ (ICLR), 2021.

[paper] [code] [Img2SceneGraph]

Spotlight Presentation

Pose Augmentation: Class-agnostic Object Pose Transformation for Object Recognition
TL;DR: Eliminate-add generator to explicitly disentangle pose from object identity by maximizing pose entropy

Yunhao Ge, Jiaping Zhao, Laurent Itti
European Conference on Computer Vision (ECCV), 2020.

[paper] [github] [video-1min] [video-10min]

Beneficial Perturbation Network for designing general adaptive artificial intelligence systems
TL;DR:Allowing a single network to learn potentially unlimited parallel input to output mappings, and to switch on the fly between them at runtime.

Shixian Wen, Amanda Rios*, Yunhao Ge* and Laurent Itti (*=equal contribution)
IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2021.


Unpaired MR to CT Synthesis with Explicit Structural Constrained Adversarial Learning

Yunhao Ge*, Dongming Wei*, Zhong Xue, Yiqiang Zhan, Xiang Zhou, Qian Wang and Shu Liao (*=equal contribution)
IEEE International Symposium on Biomedical Imaging (ISBI), 2019.

[paper] [code]

Synthesis and inpainting-based MR-CT registration for image-guided thermal ablation of liver tumors

Dongming Wei, Sahar Ahmad, Jiayu Huo, Wen Peng, Yunhao Ge, Zhong Xue, Pew-Thian Yap, Wentao Li, Dinggang Shen, Qian Wang
International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2019.


Unpaired Whole-body MR to CT Synthesis with Correlation Coefficient Constrained Adversarial Learning

Yunhao Ge, Zhong Xue, Yiqiang Zhan, Xiang Zhou and Shu Liao
SPIE-Medical Imaging (SPIE), 2019.

[paper] [code]

Oral Presentation

Intern & Work Experience

Google Cloud AI, Mountain View, USA (Aug. 2021 - current)

Microsoft Research, Redmond, USA (May. 2021 - Aug. 2021)

UII America, Inc, Boston, USA (May. 2020 - Aug. 2020)

Flexiv Ltd, Shanghai, China (May. 2019 - Aug. 2019)

United Imaging Intelligence Co., Ltd, Shanghai, China (Jun. 2018 - Apr. 2019)


Honors and Awards

Academic Service

Reviewer of the following conferences/journals:

CVPR 2022
ICLR 2022
ICML 2022
NeurIPS 2021
IEEE Transactions on Medical Imaging (TMI)
IEEE Access
Applied Optics

Software & Patents

Systems and methods for image processing
S Liao, GE Yunhao, WEI Dongming
US Patent App. 16/729,303.

Pulmonary Nodular Assisted Detection System Based on AI(V1.0)
Bin Li, Yunhao Ge

A Two-Layer Barrier Free Parking Equipment Based on Bionic Manipulator
Yunhao Ge, Shangze Yang, Zheng Zhang, Weixin Yan and Yanzheng Zhao

A Double Decker Parking Equipment based on Shear Lifting Mechanism and Hydraulic Mechanism
Yunhao Ge, Xulong Zhou, Peng Liu and Yanzheng Zhao

Last update: Jan. 21, 2022